Multiplex Detection and Quantification of Virus Co-Infections Using Label-free Surface-Enhanced Raman Spectroscopy and Deep Learning Algorithms.

Summary

A label-free SERS–deep learning platform (MultiplexCR) accurately classifies and quantifies respiratory virus coinfections directly from saliva within 15 minutes. Trained on over 1.2 million spectra across 11 viruses and mixed combinations, it reached 98.6% classification accuracy with precise concentration regression and maintained performance in blind tests.

Key Findings

• Collected >1.2 million SERS spectra from saliva across 11 viruses, nine two-virus mixtures, and four three-virus mixtures.
• Achieved 98.6% accuracy for coinfection classification and mean absolute error of 0.028 for concentration regression.
• Blind tests confirmed consistent high accuracy and precise quantification; full workflow completed in ~15 minutes.
• Utilized silica-coated silver nanorod array substrates enabling sensitive, label-free detection.

Clinical Implications

If clinically validated, this platform could enable same-visit differentiation of single versus coinfections and quantitative viral load estimation, informing isolation decisions, antiviral choices, and outbreak control.

Limitations

• Clinical validation in real patient cohorts and diverse saliva matrices is not reported
• Potential variability in SERS substrate manufacturing and matrix effects may affect robustness outside controlled settings

Future Directions

Prospective clinical validation across care settings, robustness to matrix variability, head-to-head comparisons with RT-PCR/antigen tests, and pathway to regulatory clearance for point-of-care deployment.